Learning to Find Common Objects Across Few Image Collections

Code for the paper Learning to Find Common Objects Across Few Image Collections. This is a reimplementation of the original code in TF2. The original TF1 implementation can be found here. The results might be slightly different from the paper due to the randomness.

If you use this code, please cite our paper:

@inproceedings{shaban19learning,
 author = {Shaban, Amirreza and Rahimi, Amir and Bansal, Shray and Gould, Stephen and Boots, Byron and Hartley, Richard},
 booktitle = {Proceedings of the International Conference on Computer Vision ({ICCV})},
 title = {Learning to Find Common Objects Across Few Image Collections},
 year = {2019}
}

Installation

• This code has been tested on Ubuntu 16.04 with Python 3.5.2 and Tensorflow 2.0.0.
• Install Tensorflow 2.0.
• Install EasyDict by running pip install easydict.

How to perform evaluation

• We have placed pre-trained models and config files experiments/mini/bs* directories. The config files are used to evaluate the pre-trained models. The evaluation will be performed on the test classes of the mini-ImageNet dataset.
• Run python eval.py --experiments_dir=path/to/evaluation_directory to perform evaluation. The experiments_dir argument should point to the directory where the config.json file is located.

How to train the network

• Unzip the mini-ImageNet training dataset in data/ folder. A few number of .pkl files should be located at data/miniimagenet_v2/ folder afterwards.
• We have placed config.json files for miniImageNet experiments in experiments/mini/k* directories. You can copy and edit them for your desired task.
• Run python train.py --experiments_dir=path/to/training_direcotry to start the training process. The experiments_dir argument should point to the directory where the config.json file is located.